Synthetic Data. The synthetic data represents patients' information from 2003 until June 2018 but due to the synthesis process the data will be spread beyond this range (for computational reasons only). The data contains information on approximately 190,000 synthetic patients in a distribution close to reality in Leumit. The data however, does not claim to be a full representation of the population. For a detailed explanation, please refer to The Data page at the Pay & Research website. This cohort is randomly selected and is a representative sample of the population in terms of demographic, medical condition and other aspects. The synthetic data is structured in such a way that it simulates the statistical behavior of the real population and enables a reliable analysis and a very close approximation to the real-world data. The data will be accessible in CSV format files and contains the following types of information: • Customer Details: Gender, demographics, customer status, etc. • Provider’s Treatments: Data regarding hospitalizations, ERs, and other treatments (treatment code, date, etc.) as reported by external providers. • Physician’s/Xxxxx’s Procedures: Procedures performed by the Physicians/ nurses in the HMO clinics (Procedure code, date, etc.).
Synthetic Data. To produce the bounds in Figs. 5.1 & 5.2 we created noisy mock Ly-α spec- tra (data) from exact hydrogen overdensities by adding a Gaussian white noise to the flux to simulate instrumental noise. We then use overdensities (without noise added) to fit HI assuming no dilaton coupling to the data as discussed in our pipeline in Sec. 3.3. The reconstruction (inversion) is thus noisy and we will model this in our likelihood analysis Sec. 4.3. We also simulated the dilaton effect in a wide range of masses (from 10 20 eV down to 10 32 eV) and couplings. We chose dark matter fractions that are in agreement with the combined 1σ exclusion CMB and galaxy clus- tering limits of [109, 110]. We present in Table 4.1 the masses assumed and corresponding fractions Wφ =WDM. As the broadening effect could be small, we modeled our synthetic data on high-resolution and high-quality tomographic data. In fact, we mim- icked the spectral resolution and signal-to-noise ratio (SNR) of the UVES SQUAD survey [14]. UVES SQUAD is a fully reduced spectroscopic survey of 467 QSO’s starting at very low redshift and up to redshift 5. The quasars have a median signal-to-noise ratio of 20 at a spectral resolution of 2:5 km=s at a wavelength of 5500 A˚ [14]. We used a simplified set of synthetic data where we assumed that all lines of sight are at the same redshift z = 2:5 and all spectra had the same spectral resolution 2:5 km=s and same SNR 20.
